ai-fine-tuning

Pass

Audited by Gen Agent Trust Hub on May 13, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill references additional components from the author's own repository to be installed via npx. These are legitimate vendor resources used for skill discovery and installation within the platform.
  • [PROMPT_INJECTION]: The skill establishes a workflow for ingesting external data sets, creating a surface for indirect prompt injection as defined in the analysis framework.
  • Ingestion points: Data is loaded from JSON files like labeled_data.json, tickets.json, and qa_data.json across SKILL.md and examples.md.
  • Boundary markers: Data is structured into dspy.Example objects, but the provided examples do not include explicit 'ignore instructions' boundary markers within the prompts.
  • Capability inventory: Scripts demonstrate file reading/writing (json.load, program.save) and network access for interacting with LLM providers (dspy.LM).
  • Sanitization: No specific input sanitization or validation of the training data content is implemented in the provided examples.
Audit Metadata
Risk Level
SAFE
Analyzed
May 13, 2026, 06:46 PM
Security Audit — agent-trust-hub — ai-fine-tuning